CN109389152B - Refined identification method for power transmission line falling object - Google Patents

Refined identification method for power transmission line falling object Download PDF

Info

Publication number
CN109389152B
CN109389152B CN201811004041.2A CN201811004041A CN109389152B CN 109389152 B CN109389152 B CN 109389152B CN 201811004041 A CN201811004041 A CN 201811004041A CN 109389152 B CN109389152 B CN 109389152B
Authority
CN
China
Prior art keywords
image
interest
transmission line
power transmission
region
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN201811004041.2A
Other languages
Chinese (zh)
Other versions
CN109389152A (en
Inventor
鲁仁全
王士贤
张斌
吴元清
陶杰
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201811004041.2A priority Critical patent/CN109389152B/en
Publication of CN109389152A publication Critical patent/CN109389152A/en
Application granted granted Critical
Publication of CN109389152B publication Critical patent/CN109389152B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering

Abstract

The invention discloses a method for finely identifying a pendant of a power transmission line, which comprises the following steps: reading an initial background image, modeling the initial background image according to a Gaussian mixture model, and generating a background model; reading an image to be identified, and determining an interested area by a difference method; extracting an image of the region of interest, and performing adaptive median filtering; step four, identifying the falling object by utilizing a multi-scale AdaBoost algorithm; according to the method, the region of interest is extracted by using methods such as a Gaussian mixture model, a difference method and perspective transformation, and a multi-scale AdaBoost algorithm is combined, so that the problem of poor training caused by insufficient training samples is solved, a network can learn more structural characteristics, a complex background can be effectively segmented when image data to be recognized is analyzed, and the accuracy of automatic recognition is remarkably improved.

Description

Refined identification method for power transmission line falling object
Technical Field
The invention relates to the technical field of image processing, in particular to a method for finely identifying a pendant of a power transmission line.
Background
The structure of a power system in China is complex, the distance between an energy center and a load center is long, and high-voltage long-distance power transmission is often needed. The topography and landform of the part where the overhead transmission line passes through are changeable, the natural environment of some places is severe, and in addition, the influence of wind, rain, lightning and the like easily causes dangerous conditions such as strand breakage, foreign matter adhesion and the like of the transmission line, and causes great threat to the safe and stable operation of a power system.
At present, the detection of the falling object of the power transmission line is mainly realized by manual inspection, unmanned aerial vehicle inspection or fixed point video acquisition, so that image information to be detected of the power transmission line is obtained, and then the image information is manually analyzed to make judgment and marking. The time consumed by the manual detection mode and the occupied memory are huge, the efficiency is low, and the problems cannot be found in time, so that the automatic identification detection technology needs to be researched.
In recent years, a robot vision technology is taken as a subject background of a emerging line patrol technology, a target detection technology is applied to detection of safe operation of a power transmission line, and an interested area in an image is automatically extracted by comprehensively applying various leading edge theories such as image processing, pattern recognition and a neural network system, so that a pendant on the power transmission line is detected or recognized. However, due to the reasons of complex background information of the power transmission line and various types of falling objects, the accuracy of detection and identification is not high.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a refined identification method for a pendant of a power transmission line.
The purpose of the invention is realized by the following technical scheme:
a fine identification method for a pendant of a power transmission line comprises the following steps:
reading an initial background image, modeling the initial background image according to a Gaussian mixture model, and generating a background model;
reading an image to be identified, and determining an interested area by a difference method;
extracting an image of the region of interest, and performing adaptive median filtering;
and step four, identifying the falling object by using a multi-scale AdaBoost algorithm.
Preferably, the specific process of the step one is as follows:
firstly, reading an initial background image, then modeling each pixel of the initial background image by using a mixed Gaussian model formed by K Gaussian distributions respectively, and generating a background model, namely:
Figure BDA0001783605950000021
wherein
Figure BDA0001783605950000022
An estimated value of a weight coefficient representing the ith Gaussian distribution in the mixed Gaussian model at the time t; η represents a gaussian distribution probability density function; x is the number ofj＝[xjB xjG xjR]Represents the value of pixel j at time t, xjB、xjGAnd xjRB, G, R pixel values of three channels respectively;
Figure BDA0001783605950000023
and
Figure BDA0001783605950000024
respectively representing the mean vector and the covariance matrix of the ith Gaussian distribution in the mixed Gaussian model at the time t.
Preferably, the specific process of step two is as follows:
subtracting the image to be recognized and the background model, and taking an absolute value, namely: i (x)j)-P(xj) L, wherein I (x)j) Pixel values of an image to be identified; if I (x)j)-P(xj)|>T, the pixel point is a point of interest, wherein T is an adjustable threshold; finally, the region of interest is determined by a minimum rectangle that encloses all the points of interest.
Preferably, the specific process of step three is as follows:
and segmenting the region of interest from the image to be identified by using perspective transformation, transforming the region of interest into an upright rectangular image of interest, and then performing filtering processing on the image of the region of interest by using an adaptive median filtering algorithm.
Preferably, the specific process of step four is as follows:
(1) collecting image data of a pendant of the power transmission line, and constructing a training input matrix;
(2) training a plurality of weak classifiers on a multi-scale by utilizing a training input matrix, then cascading the weak classifiers, and obtaining a strong classifier through iterative operation on the basis of updating the weights of the weak classifiers;
(3) and identifying the pendants of the images processed in the step three according to the strong classifier.
Compared with the prior art, the invention has the following beneficial effects:
according to the method, the region of interest is extracted by using methods such as a Gaussian mixture model, a difference method and perspective transformation, and a multi-scale AdaBoost algorithm is combined, so that the problem of poor training caused by insufficient training samples is solved, a network can learn more structural characteristics, a complex background can be effectively segmented when image data to be recognized is analyzed, and the accuracy of automatic recognition is remarkably improved.
Drawings
FIG. 1 is a flow chart of the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
As shown in fig. 1, a method for finely identifying a pendant of a power transmission line includes the following steps:
reading an initial background image, modeling the initial background image according to a Gaussian mixture model, and generating a background model;
the specific process of the step one is as follows:
firstly, reading an initial background image, then modeling each pixel of the initial background image by using a mixed gaussian model formed by K gaussian distributions, namely, regarding the color presented by a pixel point as a random variable Ψ, and obtaining a pixel value of a video frame image at each time T ═ 1, …, T only being a sampling value of the random variable Ψ, thereby generating a background model, that is:
Figure BDA0001783605950000041
wherein
Figure BDA0001783605950000042
An estimated value of a weight coefficient representing the ith Gaussian distribution in the mixed Gaussian model at the time t; η represents a gaussian distribution probability density function; x is the number ofj＝[xjB xjG xjR]Represents the value of pixel j at time t, xjB、xjGAnd xjRB, G, R pixel values for three channels, respectively;
Figure BDA0001783605950000043
and
Figure BDA0001783605950000044
respectively representing the mean vector and the covariance matrix of the ith Gaussian distribution in the mixed Gaussian model at the time t.
Reading an image to be identified, and marking different areas of the image to be identified and a background model within a certain threshold value by a difference method, so as to determine an interested area;
wherein the specific process of the second step is as follows:
subtracting the image to be recognized and the background model, and taking an absolute value, namely: i (x)j)-P(xj) L, wherein I (x)j) Pixel values of an image to be identified; if I (x)j)-P(xj)|>T, the pixel point is a point of interest, wherein T is an adjustable threshold; finally, the region of interest is determined by a minimum rectangle that encloses all the points of interest.
Extracting an image of the region of interest, and performing adaptive median filtering;
the third step comprises the following specific processes:
and segmenting the region of interest from the image to be identified by using perspective transformation, transforming the region of interest into an upright rectangular image of interest, and then performing filtering processing on the image of the region of interest by using an adaptive median filtering algorithm.
In particular, the perspective transformation is determined by four reference points of a rectangle in the image to be recognized and four reference points of an upright rectangular image of interest, while the four reference points of the upright rectangular image of interest are adjustable;
the transformation formula for the corresponding position of each group of perspective transformation is:
Figure BDA0001783605950000051
wherein x 'and y' are fiducial coordinates of the upright rectangular image of interest; s is a transformation scale; x and y are reference point coordinates of a rectangle in the image to be recognized;
Figure BDA0001783605950000061
is a perspective transformation matrix;
after the perspective transformation matrix is calculated, coordinates surrounded by rectangles in the image to be recognized are substituted in sequence, and then the image of interest in the upright rectangles can be obtained through transformation.
The window size S of the self-adaptive median filtering algorithm is adjustable and is [ S ]min,Smax](ii) a The adaptive median filtering algorithm comprises the following specific steps:
step A: calculating the median value Z of all pixel values in the action domain of the adaptive median filtering algorithmmed；
And B: if Z ismin<Zmed<ZmaxE, jumping to the step E;
and C: increasing the window size of the adaptive median filtering algorithm if S is less than or equal to SmaxSkipping to the step A;
step D: output Zmed；
Step E: if Z ismin<Zxy<ZmaxOutput Zxy；
Step F: output Zmed；
Wherein ZminIs the minimum value of all pixel values in the action domain of the adaptive median filtering algorithm; zmaxThe maximum value of all pixel values in the action domain of the adaptive median filtering algorithm; zxyThe value of the pixel point of the y row and the x column in the image is obtained.
And step four, training and learning the image data of the falling object of the power transmission line by using a multi-scale AdaBoost algorithm, training a plurality of weak classifiers on a plurality of image scales, and then cascading the weak classifiers to form a strong classifier, thereby finely identifying the falling object.
Wherein the specific process of the fourth step is as follows:
(1) collecting image data of a pendant of the power transmission line, and constructing a training input matrix;
(2) training a plurality of weak classifiers on a multi-scale by utilizing a training input matrix, then cascading the weak classifiers, and obtaining a strong classifier through iterative operation on the basis of updating the weights of the weak classifiers;
(3) and identifying the pendants of the images processed in the step three according to the strong classifier.
Reading an initial background image, modeling the initial background image according to a Gaussian mixture model to generate a background model, reading an image to be identified, and determining an interested area by a difference method; then extracting an image of the region of interest and carrying out self-adaptive median filtering processing; and identifying the falling object by utilizing a multi-scale AdaBoost algorithm.
According to the method, the region of interest is extracted by using methods such as a Gaussian mixture model, a difference method and perspective transformation, and a multi-scale AdaBoost algorithm is combined, so that the problem of poor training caused by insufficient training samples is solved, a network can learn more structural characteristics, a complex background can be effectively segmented when image data to be recognized is analyzed, and the accuracy of automatic recognition is remarkably improved.
The present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents and are included in the scope of the present invention.

Claims (3)

1. A fine identification method for a pendant of a power transmission line is characterized by comprising the following steps:
reading an initial background image, modeling the initial background image according to a Gaussian mixture model, and generating a background model;
reading an image to be identified, and determining an interested area by a difference method;
extracting an image of the region of interest, and performing adaptive median filtering;
step four, identifying the falling object by utilizing a multi-scale AdaBoost algorithm;
the specific process of the second step is as follows:
subtracting the image to be recognized and the background model, and taking an absolute value, namely: i (x)j)-P(xj) L, wherein I (x)j) Pixel values of an image to be identified; if I (x)j)-P(xj) If the value is greater than T, the pixel point j is an interest point, wherein T is an adjustable threshold value; finally, determining the region of interest by using a minimum rectangle which can surround all the points of interest;
the specific process of the step one is as follows:
firstly, reading an initial background image, then modeling each pixel of the initial background image by using a mixed Gaussian model formed by K Gaussian distributions respectively, and generating a background model, namely:
Figure FDA0003470649760000011
wherein
Figure FDA0003470649760000012
An estimated value of a weight coefficient representing the ith Gaussian distribution in the mixed Gaussian model at the time t; η represents a gaussian distribution probability density function; x is the number ofj＝[xjB xjG xjR]Represents the value of pixel j at time t, xjB、xjGAnd xjRB, G, R pixel values for three channels, respectively;
Figure FDA0003470649760000021
and
Figure FDA0003470649760000022
respectively representing the mean vector and the covariance matrix of the ith Gaussian distribution in the mixed Gaussian model at the time t.
2. The method for finely identifying the pendant of the power transmission line according to claim 1, wherein the specific process of the third step is as follows:
and segmenting the region of interest from the image to be identified by using perspective transformation, transforming the region of interest into an upright rectangular image of interest, and then performing filtering processing on the image of the region of interest by using an adaptive median filtering algorithm.
3. The method for finely identifying the pendant of the power transmission line according to claim 1, wherein the specific process of the fourth step is as follows:
(1) collecting image data of a pendant of the power transmission line, and constructing a training input matrix;
(2) training a plurality of weak classifiers on a multi-scale by utilizing a training input matrix, then cascading the weak classifiers, and obtaining a strong classifier through iterative operation on the basis of updating the weights of the weak classifiers;
(3) and identifying the pendants of the images processed in the step three according to the strong classifier.
CN201811004041.2A 2018-08-30 2018-08-30 Refined identification method for power transmission line falling object Active CN109389152B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201811004041.2A CN109389152B (en) 2018-08-30 2018-08-30 Refined identification method for power transmission line falling object

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201811004041.2A CN109389152B (en) 2018-08-30 2018-08-30 Refined identification method for power transmission line falling object

Publications (2)

Publication Number Publication Date
CN109389152A CN109389152A (en) 2019-02-26
CN109389152B true CN109389152B (en) 2022-05-13

Family

ID=65418689

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201811004041.2A Active CN109389152B (en) 2018-08-30 2018-08-30 Refined identification method for power transmission line falling object

Country Status (1)

Country Link
CN (1) CN109389152B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109977962B (en) * 2019-03-21 2022-03-15 国网山东省电力公司经济技术研究院 Method and system for automatically identifying fault hidden danger of optical cable
CN113486865B (en) * 2021-09-03 2022-03-11 国网江西省电力有限公司电力科学研究院 Power transmission line suspended foreign object target detection method based on deep learning

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103473791A (en) * 2013-09-10 2013-12-25 惠州学院 Method for automatically recognizing abnormal velocity event in surveillance video
CN107133974A (en) * 2017-06-02 2017-09-05 南京大学 The vehicle type classification method that Gaussian Background modeling is combined with Recognition with Recurrent Neural Network

Family Cites Families (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US8306283B2 (en) * 2009-04-21 2012-11-06 Arcsoft (Hangzhou) Multimedia Technology Co., Ltd. Focus enhancing method for portrait in digital image
US8520956B2 (en) * 2009-06-09 2013-08-27 Colorado State University Research Foundation Optimized correlation filters for signal processing
US8548256B2 (en) * 2010-07-01 2013-10-01 Intellectual Ventures Fund 83 Llc Method for fast scene matching
CN101923652B (en) * 2010-07-23 2012-05-30 华中师范大学 Pornographic picture identification method based on joint detection of skin colors and featured body parts
CN103646241B (en) * 2013-12-30 2017-01-18 中国科学院自动化研究所 Real-time taxi identification method based on embedded system
CN104463195B (en) * 2014-11-08 2018-08-07 沈阳工业大学 Printing digit recognizing method based on template matches
CN107103303A (en) * 2017-04-27 2017-08-29 昆明理工大学 A kind of pedestrian detection method based on GMM backgrounds difference and union feature

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103473791A (en) * 2013-09-10 2013-12-25 惠州学院 Method for automatically recognizing abnormal velocity event in surveillance video
CN107133974A (en) * 2017-06-02 2017-09-05 南京大学 The vehicle type classification method that Gaussian Background modeling is combined with Recognition with Recurrent Neural Network

Also Published As

Publication number Publication date
CN109389152A (en) 2019-02-26

Similar Documents

Publication Publication Date Title
CN111444821B (en) Automatic identification method for urban road signs
CN109753929B (en) High-speed rail insulator inspection image recognition method
CN107784661B (en) Transformer substation equipment infrared image classification and identification method based on region growing method
CN111428748B (en) HOG feature and SVM-based infrared image insulator identification detection method
CN109101924B (en) Machine learning-based road traffic sign identification method
CN109165623B (en) Rice disease spot detection method and system based on deep learning
CN108229550B (en) Cloud picture classification method based on multi-granularity cascade forest network
CN107038416B (en) Pedestrian detection method based on binary image improved HOG characteristics
CN104463196A (en) Video-based weather phenomenon recognition method
CN103218831A (en) Video moving target classification and identification method based on outline constraint
CN113205063A (en) Visual identification and positioning method for defects of power transmission conductor
CN109086803B (en) Deep learning and personalized factor-based haze visibility detection system and method
CN101615244A (en) Handwritten plate blank numbers automatic identifying method and recognition device
CN111898627B (en) SVM cloud microparticle optimization classification recognition method based on PCA
CN110969171A (en) Image classification model, method and application based on improved convolutional neural network
CN111855500A (en) Intelligent composite insulator aging degree detection method based on deep learning
CN109389152B (en) Refined identification method for power transmission line falling object
CN109598200B (en) Intelligent image identification system and method for molten iron tank number
Ye et al. Overhead ground wire detection by fusion global and local features and supervised learning method for a cable inspection robot
CN106570885A (en) Background modeling method based on brightness and texture fusion threshold value
CN111540203B (en) Method for adjusting green light passing time based on fast-RCNN
CN111915583A (en) Vehicle and pedestrian detection method based on vehicle-mounted thermal infrared imager in complex scene
CN113128476A (en) Low-power consumption real-time helmet detection method based on computer vision target detection
CN109409409A (en) The real-time detection method of traffic sign based on HOG+CNN
CN111191510B (en) Relation network-based remote sensing image small sample target identification method in complex scene

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant